This section repeats the wishlist of requirements from Translation Tools on the GF grammar and translation API.

Assume the GF translation goes to a reviser, working with or without another copy of the MOLTO translation tool. The corrected translation, in XLIFF form, should be brought to GF's attention. This calls for a new functionality from the GF grammar API: one which corrects the grammar and lexicon software to produce the output required by the corrected translation. This functionality is to be built on the GF example-based grammar writing methodology.

In order for the corrections to converge, revised translations must accumulate so that the newest corrections do not falsify earlier ones. The collection of manual corrections may become ambiguous or inconsistent, which situation should also be recognised and brought to the attention of a grammar engineer. Again, it is important to pay attention to user roles and rights.

We may want to provide ways to override GF translations with canned translations. At the translation tools level, this can happen by preferring TM translations over GF. We should also consider ways to override compositional translations on GF grammar level.

Another requirement is translation time update of grammar, at least the lexicon, so that translator's on the fly lexicon additions are possible.

If we want to support translating formatted documents using XLIFF, the minimum requirement is that the GF translation API handles XLIFF inline tags.